JustCall

The tier structure settles the removal test without any argument being needed. Model driven capability is packaged and sold separately rather than built in: the review and coaching layer is a $9 per user add on below Pro Plus, and the voice agent is a separate product priced per minute, from $0.99 with volume discounts down to $0.15.

Strip both away and the $29 entry tier is still a complete cloud telephone system, with call menus, routing, distribution, recording, voicemail, messaging, numbers in more than seventy countries, over a hundred record system integrations, analytics and programmatic access. That is what most of this vendor's customers are buying and it contains no inference at all.

The model layer improves a working product rather than constituting it, which is the established platform pattern this index grades in the middle of the band. Ask what the platform costs and delivers with the review add on and the voice agent both excluded.

A voice agent that answers and places real telephone calls, with a genuine escalation path and little else published. The autonomy is substantial and stated: the agent runs around the clock, and its published action set lets it collect data, schedule appointments, send messages, assign disposition codes and transfer calls, with custom actions and outbound calling at the upper plan, the latter marked as in testing.

Transferring to a person is named as a first class action, which is the escalation route this axis looks for and is more than many agent products publish. Around it sit advanced roles and permissions from the second tier and knowledge bases scoped per account.

What is not published is anything bounding behaviour: no statement of what the agent may commit to, when it must hand over, whether conversations are reviewed, what happens when it books an appointment wrongly, or what limits apply to outbound campaigns run by an agent rather than a person. Ask what the agent may not do, and what triggers a mandatory transfer to a person.

The interfaces and the metering are specific, the model layer is not disclosed. On the published side a developer reference is named and linked for programmatic access and webhooks, transcription overage is priced at a stated rate per minute, the voice agent's language support is given as a count, its rate is published per minute, and its capability set is enumerated action by action. A buyer can establish what is bought and what it costs to exceed.

On the undisclosed side nothing at all is named for transcription, scoring, sentiment, script checking, agent assist or the voice agent itself, and no provider, model family or version appears anywhere reached. The inbound receptionist can speak in a clone of the customer's own spokesperson, with no description of how a voice is captured, verified as belonging to the person cloned, or retired. Ask which models power transcription, scoring and the voice agent, and what consent and verification govern a cloned voice.

Three customer results, each attached to a named person at a named company with a case study behind it, which is the form this axis rewards. A lighting manufacturer reports dropped calls falling by thirty to forty percent, quoted by its chief growth officer by name. An education company reports a twenty five percent rise in agent productivity and around ten hours saved per account manager per week, attributed to its global operations manager.

A pharmacy reports engagement up more than twenty percent, attributed to its chief executive. Each links to a full case study, and a broader customer story section, a customer review wall and named logos across consulting, media and education sit behind them, alongside independent review standing across a large review base. It is held below the top band by the measurement basis, absent in all three cases: no period, no baseline, no method and no definition of the metrics. A separate claim of powering more than a billion conversations carries no date or counting basis at all. Ask over what period each figure was measured and against what baseline.

Registration obligations are handled and the dialing regulation this product most engages is never mentioned. On the credit side, messaging registration for the United States is acknowledged as a real cost with compliance charges described in the help material, script checking is shipped as a named feature at the upper tiers, a number health capability exists on the outbound package, and the fair usage policy sets published numeric ceilings.

Against that, the product sells a predictive dialer running up to ten lines per agent, which is the architecture that abandoned call regulation exists to govern, and nothing on any page reached states an abandonment rate ceiling, an answering machine detection policy, a calling window restriction or a screening step against do not call registers.

Two records in this lane hold the top grade for exactly those things, one for engaging the regulation by name and one for putting register screening in its cheapest plan. Ask what abandonment rate the predictive dialer enforces and whether numbers are screened against do not call registers before dialing.

The instruments are published, were not read, and the data protection terms are sold rather than provided. A privacy policy, terms and conditions, a responsible disclosure policy and a trust centre are all published, and none was retrievable on this pass, so retention periods, lawful bases and the rights exercise route are unestablished and nothing here asserts they are absent. What the pricing page does establish is a gating decision worth recording.

Enterprise grade compliance covering the health information regime and European data protection is listed as a feature of the quoted tier with a ten licence minimum, which places contractual data protection terms behind a negotiated purchase rather than offering them to every customer. That matters because this product records telephone calls and stores message content involving people who are not customers, on every tier. Ask what data protection terms apply on the published tiers, and what the default retention period is for call recordings and message history.

No purchased corpus, and the material at stake arrives from the customer and the network. There is no contact database, no enrichment and nothing brokered here: the numbers dialed come from the customer's own record system, and what the platform accumulates is call recordings, transcripts and message history generated by use.

So the usual chain of sources, licences and resale permissions does not apply and most of the axis is out of scope, which is why this sits in the middle of the band with the scope stated. What remains in scope is that recordings and transcripts contain the speech of the people called, who are not customers of either party and whose consent to recording is a legal precondition in many jurisdictions.

Nothing published describes how the platform supports, prompts for or evidences that consent, or whether an announcement can be enforced at the account level. Ask whether a recording announcement can be made mandatory per jurisdiction and what consent record is retained.

The telephony is the vendor's own and the software connections run through sanctioned routes. Numbers are provisioned by the vendor in more than seventy countries and calls and messages traverse carrier infrastructure the vendor contracts for, so a customer is not borrowing access to a platform that could withdraw it, and porting numbers in and out is offered as a supported process.

The integration surface reaches customer record, support and sales engagement systems through each product's documented route, with more than a hundred named and a public developer reference for anything custom, plus mobile applications distributed through both official stores. That architecture is what this axis rewards.

It stops below the top band because the governing regime here is carrier and messaging registry rules rather than software platform terms, and while messaging registration is acknowledged as a cost, no conformance position of the vendor's own is stated against carrier requirements or the consequences of a campaign registration being revoked.

The load bearing commitment for a product of this shape is not stated anywhere reached. This platform ingests recorded telephone calls and message content and runs transcription, scoring, sentiment analysis and coaching over them, which makes the training question direct: do a customer's recorded conversations train or tune the models, and does a third party model provider retain them.

No statement on either appears on the pricing page, in the questions section or in the navigation, where a comparable vendor graded this session places its position on its own pricing page. The trust centre was not retrievable on this pass, so nothing here asserts the commitment is absent, only that it was not found where a buyer would look. The receptionist's cloned spokesperson voice raises a second unaddressed question about what is retained of a cloned voice and who may use it. Ask for a written statement that recordings do not train any model, and what a cloned voice sample is retained for and for how long.

Outbound calls from JustCall's voice agent are sold only for contacts who gave prior express consent to AI or prerecorded calls, and JustCall rules out cold calling and purchased lists for the agent. JustCall describes a compliant outbound call as one that opens with clear identification and an AI disclosure, offers an easy opt out during the call and runs only in permitted hours, and leaves the customer's legal team to confirm the setup before campaigns scale.

That opening is part of the call flow the customer builds, not a line the platform inserts or enforces. The inbound receptionist has no matching commitment: it speaks in professionally recorded voices or a clone of the customer's own spokesperson, and JustCall concedes that no AI voice sounds human in every situation while adding that most callers won't notice. Article 50 of the EU AI Act goes unnamed, though the EU appears among the regions whose calling rules a compliant setup follows. Recording runs on location based consent rules, so in all party consent jurisdictions the person on the line is asked before recording starts or the call is not recorded.

Programmatic access on the entry tier and a stated count backed by named destinations. More than a hundred integrations are claimed and fifteen are named individually with their own pages, spanning three customer record platforms, two support desks, three sales engagement tools, a scheduling product, a messaging platform, a prospecting database and a general automation connector, so the claim is checkable rather than a logo wall.

Interface access and workflows appear in the entry plan's feature list rather than being reserved for the largest customers, with higher limits at the second tier and maximum limits at the quoted tier, and a public developer reference is named and linked in the questions section. Mobile applications ship on both official stores, and an integration with a vehicle interface exists.

It stops below the top band because the developer reference was not retrieved on this pass so what the interface exposes is unverified, and because rate limits are tiered in a way the page does not quantify at any level.

Telephony geography is documented in unusual detail and data geography is not addressed. The calling side is specific: numbers in more than seventy countries, a region selector on the pricing page covering around eighty countries that changes the published rates, local number bundles available in seven named markets, and per region inclusion and exclusion lists stating exactly which call routes and message destinations fall inside a bundle.

A buyer can establish where they can call and what it costs with more precision than almost any record in this lane. What is absent is where anything is stored. No hosting region, processing location, residency election or cross border transfer mechanism for recordings, transcripts or message history appears on any page reached, and the trust centre was not retrievable.

That gap matters because the stored material is recorded speech from whichever of those eighty countries a customer dials. Ask in which regions recordings and transcripts are stored and whether a region can be elected.

JustCall's pricing and security pages display six compliance marks, SOC 2, HIPAA, ISO, PCI DSS, GDPR and CSA STAR Level One, and the security page states SOC 2 Type 2 and ISO 27001:2022 certification, with a copy of the ISO certificate in its trust center, describes encryption and access controls, and offers a signable data processing addendum under GDPR.

The pricing page's FAQ still describes a SOC 2 Type 1 report available on request, so the vendor's own pages name two attestation types, and no auditor or audit period is shown publicly. A responsible disclosure policy is linked from every page. Single sign on and enterprise grade compliance covering HIPAA and GDPR belong to the quoted Business plan with a 10 license minimum.

JustCall publishes Team, Pro and Pro Plus at $29, $49 and $89 per user a month billed annually or $39, $69 and $109 monthly in the United States and Canada, each with a 2 license minimum, at rates that change across around 80 calling regions.

The cards list unlimited outbound and inbound minutes and unlimited transcription, and a fair usage policy on the same page sets the boundary: calls within the chosen region and inbound routes under $0.02 a minute are covered, while toll free, premium rate, international and forwarded calls and automated texts are not. The add ons are priced, with AI Review Assist at $9 a user, local number bundles from $150 a month and the voice agent from $0.99 a minute down to $0.15 at volume. Business is quoted, traffic outside the fair usage boundary has no rate on the pricing page, and a pricing FAQ still describes a capped transcription allowance on one regional bundle.

The asset most likely to lock a customer in can be taken away, and the records built on it are undescribed. Telephone numbers port both in and out, with the vendor stating it handles inbound porting without service interruption and publishing porting fees as a real cost, which is the single most important exit right for a telephone system because a business number is an operational dependency rather than a data file. Programmatic access on every tier provides a technical retrieval path.

What is absent is everything about the data. No export route is described for call recordings, transcripts, message history, scoring or analytics, and no statement exists on what happens to any of it after cancellation or how long a departing customer has to retrieve it. The one published deletion statement concerns lapsed trials, where the account is stated to be deleted thirty days after the trial ends. Ask for a documented export of recordings, transcripts and message history, and the retention window after a paying account closes.

Real mechanisms for the voice equivalent of deliverability, and no measurement of any of them. What is published is more than most in this lane: a number health capability on the outbound package, local number bundles in seven markets sold on the reasoning that a local presence improves answer rates, and a stated allowance of one free swap per month for numbers that are underperforming, which is a direct and practical answer to numbers being flagged or spam labelled by carriers.

Messaging registration for the United States is acknowledged as a precondition with fees. Fair usage ceilings are published as numbers per user per region. Against that, the vendor claims industry leading connect rates with no figure, no answer rate, no spam labelling incidence and no message delivery rate is published anywhere, and nothing describes how the predictive dialer paces against answer detection. Ask for measured connect and answer rates by region, and how often numbers are flagged and swapped.

Coverage documented across three dimensions rather than asserted on one. Seven industries carry their own pages, spanning healthcare, education, solar, fitness and wellness, real estate, financial services and retail, and three functional buyers are addressed separately across sales, support and technology operations, which is broader vertical treatment than any record graded in this batch.

Geographic reach is evidenced concretely rather than claimed, with numbers in more than seventy countries, a pricing region selector covering around eighty, and local number bundles named for seven specific markets. The commercial ladder runs from a two licence entry tier through two further published tiers to two quoted arrangements, one of them requiring ten licences, so small teams and large operations both have a described path.

It stops below the top band because no company size guidance accompanies the tiers beyond licence minimums, and because support coverage and service hours are not stated for any region outside the quoted tier's dedicated line.